Re: New RDF model theory

>Some minor points:

Thanks for the input.

>well-foundedness
>  I think that any output document should ensure that a reader is likely
>to come across (some note about) the justification for sets that contain
>themselves before an example. A simple way to achieve this is to have a
>subsection entitled "The WellFoundedness of the RDF Model" visible in
>the table of contents, containing some of Pat's comments.
>On first reading there was a period when I was half expecting a
>non-standard set theory with an axiom of anti-foundation. [Menzel's neat
>trick is reminiscent, if I remember correctly, of some treatment of
>anti-foundation that I saw well over a decade ago, so I am probably
>misremembering, Peter Aczel's Anti-Foundation Axiom is the one I
>explored at the time].
>

Will do.

The history of this 'trick', I have since discovered, is that Chris M 
is an old friend and colleague of Ed Zalta (they were in grad school 
together) and that indeed Aczel's mind was active somewhere in the 
background, as was apparently Dana Scott, who Ed also partly credits 
for this idea. The explicit use of the thing/extension distinction 
and a mapping between them is partly inspiredin Chris by a 
philosophical taste for nonextentionalism, but apparently arose as a 
technical device which Aczel used before grasping the anti-foundation 
nettle.  Ed has been using a similar device in his evolving 'abstract 
entity logic' (I think I have the name right) for similar reasons.

Oddly enough, I came to this stuff exactly the wrong way round, by 
first reading Aczel, then trying to convince Chris that it was the 
most suitable basis for a model theory, then being persuaded that 
this device was less controversial but could do all that was 
required. I still prefer the direct use of the Aczel theory, but the 
world doesn't seem to be ready for that yet.

><unnamed point>
>From intro to section 3: "the inference of (exists (?x) (foo (?x)) from
>(foo baz)" Not if baz is a literal and not a resource (if there are any
>such things :-)).

Right, I noticed that bug later. Now fixed.

>Probably doesn't matter the more formal stuff is
>clear. Perhaps we need to have clear separation or labelling of
>normative and non-normative material.  [I have jumped the gun, and I am
>imagining that much of this text will end up in the final documents!]

Much of the current text is written for the WG and will be taken out 
from anything normative. Also I will try to keep a better separation 
between the MT, intro/explanatory material, and 'further discussion' 
which includes all the stuff that is in fact proof theory.

>
>xml:lang
>  It would not cost very much to indicate how xml:lang can be treated in
>section 1.
>e.g.
>For serializations that permit language tagging from a set of language
>tags LT (which we will take as containing "" for untagged items, cf RFC
>3066), there is a fixed mapping XTL: LT x qLiterals to LV.

We could do that, but I would prefer to keep the MT attached to the 
RDF graph itself, and keep 'semantic' issues separate from issues of 
how to map from serializations to the graph. That keeps a lot of 
issues sharply clear that otherwise seem to generate a lot of heat 
and no light.

Pat Hayes

---------------------------------------------------------------------
(650)859 6569 w
(650)494 3973 h (until September)
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Monday, 27 August 2001 14:38:12 UTC